Publication bias in meta-analysis: its causes and consequences

Abstract

Publication bias is a widespread problem that may seriously distort attempts to estimate the effect under investigation. The literature is reviewed to determine features of the design and execution of both single studies and meta-analyses leading to publication bias, and the role the author, journal editor, and reviewer play in selecting studies for publication. Methods of detecting, correcting for, and preventing publication bias are reviewed. The design of the meta-analysis itself, and the studies included in it, are shown to be important among a number of sources of publication bias. Various factors influence an author's decision to submit results for publication. Journal editors and reviewers are crucial in deciding which studies to publish. Various methods proposed for detecting and correcting for publication bias, though useful, all have limitations. However, prevention of publication bias by registering every trial undertaken or publishing all studies is an ideal that is hard to achieve.